@kellogh@hachyderm.io
@kellogh@hachyderm.io avatar

kellogh

@kellogh@hachyderm.io

I'm a software engineer and sometimes manager. Currently #Raleigh but also #Seattle. Building ML platform for a healthcare startup. Previously, built an IoT platform for one of "those" companies.

Open source: dura, fossil, Jump-Location, Moq.AutoMock, others

Do I have other interests? No, but I do have kids and they have interests. I think that counts for something. I can braid hair and hunt unicorns!

I put the #rust in frustrate

He/Him

#metal #science #python

This profile is from a federated server and may be incomplete. Browse more on the original instance.

marcslove, to random

Can’t wait till we figure out how to communicate with marine animals in their language thanks to AI.

kellogh,
@kellogh@hachyderm.io avatar

@marcslove are you referring to the research a few weeks ago where they used ML to decipher whale calls?

kellogh, to random
@kellogh@hachyderm.io avatar

is stacked cooking a thing?

i’ve been doing stuff like steaming broccoli on top of the pan i’m cooking rice in. they both turn out well and i figure i’m recycling steam.

could i stack it higher? do other people do this?

kellogh, to random
@kellogh@hachyderm.io avatar

imagine having so little privacy that your every bladder issue is front page national news

ErikJonker, to ai
@ErikJonker@mastodon.social avatar

It seems that Google is failing again with the first attempts of integrating AI into search or are we just seeing the terrible mistakes and is 98% of the experience great ?
In marketing/PR terms , things are not going well i think. For OpenAI it's easier they don't have a userbase with billions of users (Gmail, Search, Drive etc.).

kellogh,
@kellogh@hachyderm.io avatar

@ErikJonker yeah, the trustworthiness of LLMs isn’t really an issue. the problem is their profile of trustworthiness didn’t match Google’s existing product

when you create a new product, e.g. perplexity, it’s all new so any feature is positive. when you change an existing product, each feature is a diff from the previous state. so in this case, LLMs made google decidedly worse

if they had launched a totally new product, it probably would have had a mostly positive response

kellogh, to LLMs
@kellogh@hachyderm.io avatar

i’m very excited about the interpretability work that has been doing with .

in this paper, they used classical machine learning algorithms to discover concepts. if a concept like “golden gate bridge” is present in the text, then they discover the associated pattern of neuron activations.

this means that you can monitor LLM responses for concepts and behaviors, like “illicit behavior” or “fart jokes”

https://www.anthropic.com/research/mapping-mind-language-model

kellogh,
@kellogh@hachyderm.io avatar

further, you can also artificially activate these concepts

they have a version of #claude with the “golden gate bridge” concept artificially activated, and so it tries to make everything it says about thr golden gate bridge

https://www.anthropic.com/news/golden-gate-claude

kellogh,
@kellogh@hachyderm.io avatar

so now we have a way to interpret and query #LLM responses in a structured format, as well as a control mechanism for driving LLM behavior

this is great news

Bruce Schneier wrote that prompt injection boils down to the fact that data and code pass through the same channel. with this interpretability work, we’re seeing the beginnings of a control channel separated from the data channel — you can control LLM behavior in a way that you can’t override via the data channel

https://www.schneier.com/blog/archives/2024/05/llms-data-control-path-insecurity.html

kellogh,
@kellogh@hachyderm.io avatar

this is great work. i’m excited to see where this goes next

i hope exposes this via their API. at this point in time, most of the promising interpretability work is only available on open source models that you can run yourself. it would be great to also have them available from vendors

Lobrien, to random

Nothing can be done to stop this, says only industry where this regularly happens. https://mastodon.world/@hn100/112493477923174205

kellogh,
@kellogh@hachyderm.io avatar

@Lobrien i can’t get past the toots where AI summary is citing Reddit. its like a scene from Silicon Valley (HBO)

lzg, to random
@lzg@mastodon.social avatar

frustrated once again googling what the fuck is a red herring, an idiom that never sticks to my brain no matter how many times i’ve looked it up

kellogh,
@kellogh@hachyderm.io avatar

@lzg you probably get most idioms, this one is just a red herring

kellogh, to ai
@kellogh@hachyderm.io avatar

iTerm2 developer caves to the bullies and moves the feature to a plugin

https://news.ycombinator.com/item?id=40458135

kellogh,
@kellogh@hachyderm.io avatar

@capraobscura thanks for illustrating my point

kellogh,
@kellogh@hachyderm.io avatar

@Xoriff eh, the hacker news & mastodon comments got into the bullying range pretty fast.

a lot of people seem to feel entitled to free software being catered to their wishes. i’ve run into the same sort of entitlement in software i’ve open sourced

kellogh,
@kellogh@hachyderm.io avatar

@sanityinc the whole fiasco highlights how much we demand from open source, how little respect maintainers get, and how tiny the communities are. most people didn’t even realize this was an open source project

kellogh, to LLMs
@kellogh@hachyderm.io avatar

if i had more time, i'd love to investigate PII coming from . i've seen it generate phone numbers and secrets, but i wonder if these are real or not. i imagine you could look at the logits to figure out if phone number digits were randomly chosen or if the sequence is meaningful to the LLM. anyone aware of researchers who have already done this?

kellogh,
@kellogh@hachyderm.io avatar

i would guess that phone numbers are probably mostly random, since so many phone numbers are found online, whereas AWS keys are less common, so you're probably more likely to get partial or even full real keys

lzg, to random
@lzg@mastodon.social avatar

i’ve realized i have about 5 scripted things i say to neighbors on morning walks, but they are all slightly odd and maybe need a little workshopping

kellogh,
@kellogh@hachyderm.io avatar

@lzg what? are you thinking of eating him?!

paul, to random
@paul@oldfriends.live avatar

I was looking for a project that would let me manage my Mastodon follows & followers better. Haven't found anything but did come across Mastodon+Steampipe. If you know of something, @ me

Anyone using Mastodon + Steampipe? Steampipe is an open-source zero-ETL engine to instantly query cloud APIs using SQL
https://hub.steampipe.io/plugins/turbot/mastodon

It has many different "recipes" to see the data on your instance/Mastodon account. https://hub.steampipe.io/plugins/turbot/mastodon/tables

Users &Admins can create access token and use it

kellogh,
@kellogh@hachyderm.io avatar

@paul @jcrabapple holy hell, steampipe…, i’ve needed something like this for forever

kellogh, to random
@kellogh@hachyderm.io avatar

“One House Republican called the incident "vile" and said it has caused concern among GOP lawmakers.”

anonymous coward, show thyself!

https://www.axios.com/2024/05/22/rnc-vials-blood-capitol-police-suspicious

kellogh, to random
@kellogh@hachyderm.io avatar

this has been bugging me a lot. like, yeah, there’s definitely AI scams out there. and yeah, a lot of people are using it from the wrong end, but it’s also clearly a substantial technology. time to realize that
https://mas.to/@carnage4life/112484753548884371

kellogh,
@kellogh@hachyderm.io avatar

most of the complaints about AI at this point are people using it from the wrong end and exclaiming, “see? it doesn’t work”. there are legitimate problems, ofc, but there’s also legitimate value

kellogh,
@kellogh@hachyderm.io avatar

my take on the bubble — there will definitely be some sort of decline at some point, but it’s not going to be a bubble pop as widely predicted. ML has been generally growing for 10 straight years, at an accelerating pace, also for 10 straight years. to predict a bubble pop is to ignore a whole lot of data, including the idea of AI is basically the culmination of computing in general, since its inception. it’s quite a different case from blockchain.

kellogh,
@kellogh@hachyderm.io avatar

@leoncowle one phenomenon, it sometimes seems like the anti-AI activity starts to feel more scammy than the AI applications they criticize. FUD, but with a moral sense of urgency and inconsistent logic. it sets off the scam alert in my brain. i’m not sure anyone is really making money off the anti-AI dialog, but it triggers that same pattern in my brain

kellogh,
@kellogh@hachyderm.io avatar

@maltimore i wish you weren’t right

kellogh,
@kellogh@hachyderm.io avatar

@swiftcoder i think the iTerm2 overreaction really highlighted that people have indeed stopped thinking and are instead using their emotions

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • megavids
  • rosin
  • ngwrru68w68
  • osvaldo12
  • DreamBathrooms
  • mdbf
  • magazineikmin
  • thenastyranch
  • Youngstown
  • khanakhh
  • everett
  • slotface
  • tacticalgear
  • kavyap
  • JUstTest
  • cubers
  • normalnudes
  • modclub
  • ethstaker
  • InstantRegret
  • GTA5RPClips
  • cisconetworking
  • Durango
  • Leos
  • provamag3
  • tester
  • anitta
  • lostlight
  • All magazines
    •